On May 6, 2026, OpenAI launched B2B Signals, a new metric series that visualizes how enterprises are actually using AI[1]. The initiative is the business-facing extension of OpenAI Signals, drawing on anonymized and aggregated usage data from Enterprise, Business, and API customers to publish recurring views of depth of use, agentic adoption, and industry-level patterns. The headline figure from the inaugural report: the gap between "frontier firms" and "typical firms" widened from 2x to 3.5x in just one year[1][2].

The frontier gap is about depth, not volume

OpenAI converts token generation into a per-worker measure, then defines firms at the 95th percentile as the frontier and those at the 50th percentile as typical[1]. By that yardstick, frontier firms now consume 3.5x the intelligence per worker that typical firms do, up from 2x in April 2025[1][2].

The breakdown of that gap is what stands out. OpenAI reports that raw message volume explains only 36 percent of the difference, with the remainder coming from depth—how complex each query is, how much context it carries, and how substantive the resulting output ends up being[1]. In other words, what separates leaders is no longer how many keystrokes their employees send, but the operational maturity around what they delegate to AI, how much information they hand over, and what kind of outputs they pull back.

Agentic tools widen the gap further

Looking at adoption by tool type, the largest gap by far appears with Codex. Frontier firms send 16x as many Codex messages per worker as typical firms[1][3]. The same directional pattern shows up across ChatGPT Agent, Apps in ChatGPT, Deep Research, and GPTs—the tools designed to delegate multi-step tasks or apply company-specific context. Teams that have shifted from "AI as a chat assistant" to "AI as a delegated worker" are using these capabilities far more deeply[1].

Industry breakdowns are also published. Professional, Scientific, and Technical Services ranks first in both Codex adoption and API intensity. Finance and Insurance leads in ChatGPT adoption, driven by large-scale enterprise deployments. Educational Services tops per-user message intensity[1]. Among job functions, IT and Security teams lean heavily on procedural and how-to queries, Software Development and Data Science teams focus on coding, and Finance teams concentrate on analysis and calculation—AI is moving beyond generic productivity and into the core responsibilities of each function[1].

Customer examples and privacy design

For real-world Codex deployments, OpenAI cites Virgin Atlantic (expanding test coverage and reducing technical debt), Ramp (accelerating code review), Notion (shipping new features faster), Cisco (navigating large repositories), and Rakuten (incident response)[1][3]. The message: enterprise Codex adoption is now spreading beyond engineering into adjacent functions in measurable ways.

On the data-handling side, the analysis is restricted to anonymized and aggregated usage signals. Message contents are processed by automated classifiers, and OpenAI staff do not view individual enterprise, business, or API customer data for analysis purposes[4]. Published aggregates also have calibrated noise added through differential privacy, so the presence or absence of any single individual or small group does not meaningfully change the result[4]. The framing is meant to satisfy two simultaneous needs: enterprises wanting an industry benchmark, and those same enterprises not wanting their proprietary data inspected.

Conclusion

What B2B Signals makes visible is that the AI competition between firms has moved from "who has access" to "who knows how to use it." Frontier-firm advantage is not built on raw activity—it sits on operational design: delegating complex work to agents, providing rich context, and integrating outputs into actual workflows. As the next reports push deeper into industry and function-level slicing, the series should become a useful reference point for any organization trying to set its own benchmarks.

Source: https://openai.com/index/introducing-b2b-signals/

Source: https://openai.com/signals/b2b/

Source: https://openai.com/index/scaling-codex-to-enterprises-worldwide/

Source: https://openai.com/signals/data/